| Long-range modeling of a precipitation-runoff process has become indispensable to predict/forecast runoff and study the impact of modern anthropogenic factors and land change use on watersheds. The purpose of this thesis research is to interpret, model and forecast complex drainage basins using advanced signal processing technique and a physically-based low-dimensional dynamic model.; The first emphasis is placed on a hydrogeologic interpretation of a complex drainage basin. The space-time patterns of annual, interannual, and decadal components of precipitation, temperature, and runoff (P-T-R) using long-record time series across the steep topographic gradient of the Wasatch Front in northern Utah, are examined. The singular spectrum analysis is used to detect dominant oscillations and spatial patterns in the data and to discuss the relation to the unique mountain and basin hydrologic setting. For precipitation and temperature, only the annual/seasonal spectral peaks were found to be significantly different from the underlying noise floor. Spectral peaks in runoff show increasing low-frequency components at intermediate and low elevation. A conceptual hydrogeologic model for the mountain and basin system proposes how losing streams and deep upwelling groundwater in the alluvial aquifer could explain the strong low-frequency component in streams. The research shows that weak interannual and decadal oscillations in the climate signal are strengthened where groundwater discharge dominates streamflow.; The second emphasis is focused on developing a long-range physically-based precipitation-runoff model. A low-dimensional integral-balance model is developed for a hydrologic system where multiple time scales of basin storage play the dominant role on a precipitation-runoff process. The genetic algorithm (GA) technique is implemented for parameter identification with the observed data. The model is developed for the Upper West Branch of the Susquehanna River in Pennsylvania, within the Appalachian Plateaus. Model performance was assessed for runoff over calibration and verification. The “optimal” conceptual model has two nonlinear modes: fast and slow responses. The accuracy of the model suggests the utility of low-dimensional models for probabilistic flood and drought forecasting, as well as quantifying the impacts of changing land use and climate. |
